Thank you very much. And then we're gonna move to our third talk given by Kathleen Looms, who's a... Oh, sorry, Elliot Tapper, who's at the University of Michigan. Thank you. It's my privilege to be here with you today to talk about the practical application of non-invasive tests. We're going to talk a lot about probability, we're going to go deep on some math, and the question is why. Our aim in the application of these tests is not to provide a specific stage or a diagnosis in the cross-section in the clinic as we face our patients, but rather to think longitudinally, to try to understand the probability that their liver could get sick so that we can improve outcomes by reducing costs, by making sure that the people who need a hepatologist see a hepatologist and not the converse, by reducing false positives, false negatives, while also improving effectiveness by using these tests to help us link people to the kind of care that has the ability to bend the natural history of their disease or capture bad complications before, when we have an opportunity to intervene successfully. And in order to do that, we have to personalize our approach. We have to right-size our application to the patients in front of us, and this requires understanding what is at stake for this patient. And in fact, in the average clinical visits that we have over the course of our busy days, the clinical stakes vary wildly. They're very different for the 16-year-old who's presenting to us with elevated liver enzymes as they would be for the 80-year-old presenting with the same. Our concern and the amount of confidence that we have about what particular stage they have or what the future might hold will be highly variable. How full we need to fill the cup of clinical certainty changes based on what is at stake for these patients. So take, for example, the average person who presents to us in clinic. This person with incidentally detected elevated liver enzymes has a pretest probability of zero to five percent advanced fibrosis. And this forces two questions. Number one, how do we apply the available tools in such a way that we get to a place that we're sufficiently confident to move forward and improve the outcomes to meet their unmet needs? And then two, just how certain do we have to be? So for that 16-year-old, we might be more interested in getting scared, being worried, embracing the possibility of a false positive so that we can marshal the resources that they need. But it would be very different for an 80-year-old. So how do we select the various tools available from non-invasive tests to determine what could potentially happen to our patient? Now I could spend forever reviewing the non-invasive tests that have proliferated. It seems that with each liver meeting or each issue of hepatology, a new one is added. In general, what I would rather do is discuss an approach. I'll go deep on a few of these. And along the way, we're going to talk about the best way that as clinicians we can evaluate these potential tests. So when you hear about a non-invasive test, you often see statistical descriptions of their accuracy, their sensitivity, their specificity. But the answer to which test we use in which situation is really in the form of a question. Accuracy, sensitivity, specificity, for what? And what I'm arguing is that it is not to tell someone that they have F2 or F3. It's rather to figure out how we can predict the probability of those adverse events. And there are many reasons why we're not going to refer back to a liver biopsy when we see a patient in clinical practice. There are at least two main categories of reasons. The first is that although it is often repeated that the biopsy is a gold standard, in fact, it has wide confidence intervals around each of the stages that are read to us in that dichotomous pathology report. Those reasons are three. Number one, give the same sample to two of the best hepatopathologists, and you may get a very different answer. This is very true for stages of fibrosis less than cirrhosis. It is definitely true for inflammation, and it is nearly always true for grades of steatosis. Take two different biopsies from the same patient, and you can get two different stages. And now that we've outsourced most biopsies to practitioners that use devices that yield specimens with lengths no greater than two centimeters, we simply aren't confident, as we used to be, in just how accurate those samples might be. But the second reason is that in a practice that is driven by noninvasive tests, we simply cannot provide a stage. So let's go deep on a couple of these popular tests. The first is FIB4. So this is the canonical noninvasive test using blood-based biomarkers. It's recommended by most societies as a first pass to screen for advanced fibrosis. And there are a couple of notable features that tell us why there are some patients why this may not provide us the information that we seek. First, age is part of the calculation. So for a 16-year-old, it just wasn't designed for those patients. Similarly, for a 100-year-old, it wasn't designed for those patients. But even within the sweet spot, for people between the ages of 35 and 65, it has an imperfect sensitivity and specificity for the presence of advanced fibrosis. The same is also true for elastography. So this is an individual patient meta-analysis from the best centers evaluating the performance of elastography vis-a-vis a liver biopsy gold standard. And the boxes on the left-hand side of the screen highlight the different buckets that you could put a patient into, any fibrosis, one or greater, all the way down to F4 alone. And in the middle, you can see the cutoff ranges as they vary between studies. And the key takeaway here is that for a patient with a liver stiffness measure by FibroScan of 7.9, which nobody would say is a viable cutoff for advanced fibrosis, that person could potentially have cirrhosis on their biopsy or F1 or even F0 fibrosis. The same is true for a person with a magnetic resonance elastography-derived liver stiffness measure of 3. It is often said that MRE is more accurate. It is a way to break the tie, if you will, between a Fib4 and a FibroScan. But here, if you look at this updated meta-analysis, again, from the highest volume MRE centers, if you go from panel A down through panel D, it's that same schema, buckets of fibrosis, one or greater, all the way down to just people with F4 fibrosis. And what you can see are two key things. Even though this slide is hard to fully visualize, you can see that the blue boxes are all over the place, which is to say that every center will give you different cutoffs. And even when you pool the results, you'll still have sufficient overlap between stages that we just cannot be confident in the moment that we are providing somebody with an accurate stage. So if a stage of fibrosis is not our target, what are we really interested in? We are interested in the probability that our patient will get sick. And fortunately, we have time and time again learned that it is really only those people who, if they had a liver biopsy, we found F3 or F4 fibrosis, who are on the clock for liver-related events. It is these people who need a hepatologist, who need semi-annual or annual follow-up with you. And fortunately, if we were to just bucket these stages and condition our non-invasive tests on those, we actually do a pretty good job of identifying who is at risk. So if you simply look at the people who have the highest FIB4 scores, these are the people who will go on to develop liver-related events. We've learned this in the United States, as we've seen in Europe. And on the bottom side, the slide that starts with a D, this is from the UK Biobank. These are unselected people, where it is only people who have the top percentile of FIB4, usually far in excess of 2.67, who go on to develop liver-related events. Of course, most of them will not, but it is within that group that we start to worry. On the bottom right, you see a graph here from Drs. Shearer and Ian Rowe, where they take people who had a known elevated liver stiffness and then look at the FIB4 again. And it is these people who have the highest levels, far in excess of 2.67, who develop their liver-related events. And so if we start to think more in terms of the probability of developing a sick liver, of developing a decompensation, we can move beyond focusing on specific cutoffs. We can uncap and undichotomize our view of these non-invasive tests. So here again is a great slide from Drs. Shearer and Rowe, where they look at the people across their practice at the University of Leeds in the United Kingdom. And it is the people with a liver stiffness greater than 25, sometimes 2x what we would conventionally hold to be the diagnosis for dichotomous cirrhosis. It is these people who develop liver-related events in the short to medium term. And so although this slide may appear small, it is, to me, the most important one. So the same authors, Shearer and Rowe, they take the data about the natural history of the disease, and they provide to you for the same person, age 55, changing whether or not they're a man or a woman, or have alcohol-related liver disease, or mascled down in the middle. And if you take a look there, you can see that a woman who is age 55 with a liver stiffness of 15 and a low FIB4 has a probability of decompensating at 5 years of 2.5% and a probability of liver cancer of 0.6%. This is the kind of information that people need to know. On the converse, take a look just two rows below that, a woman with a liver stiffness of 30, far in excess of conventional cut-offs for cirrhosis. Her risk of decompensation has doubled. Her risk of liver cancer has nearly doubled. So how do we apply this to the average person who we're seeing in clinic? And I would argue that clinically, there's a three-step process by which we would use these noninvasive tests to right-size our recommendations for the people in front of us. The first step is to decide just how certain we need to be. And I think for the 16-year-old, where we're all pretty worried about her, it doesn't really matter if she has a false positive, because we need everybody involved in her care to be worried and marshal the resources, and understanding that we can always revisit this in years to come. But for the average person, again, in middle age, presenting with a pretest probability of 0 to 5%, it's worth us considering just how the math plays out as we try to apply these noninvasive tests so that we can understand how confident we should be with the noninvasive tests that are available. So here's the math that I promised. This is basically an application of a Bayesian theorem. So assume at the outset that this is an average person presenting with about a 2% risk of harboring advanced fibrosis or cirrhosis without knowing it. And then we do what we're told to do, which is apply the FIB4. If that person is aged 45, the positive likelihood ratio is astronomical. And immediately, we are told that this will raise the protest probability of advanced fibrosis to 67%. The majority of people this age with a positive FIB4 are going to have advanced fibrosis, are at risk for developing liver-related events. But then for a 65-year-old, because age is part of FIB4, the positive likelihood ratio is much lower. And it only increases that probability to 7%. And so you can take a test, which we are told is far more accurate or far more specific, like magnetic resonance elastography with a liver stiffness of 3.6, or a fiber scan with a liver stiffness of 10, and that will only bring you to a place where your post-test probability is 24% or 42%. And what this means is that either you have to be comfortable with the idea that you are potentially over-diagnosing a patient with cirrhosis. And that might be OK if you're worried and you want them to be engaged in their care, or if you simply think that you can discharge them for care and get an ultrasound every six months with their primary care doctor. But it might not be OK if this person is about to apply for life insurance or the like. You have to think about what additional tests. Does this person need a lifestyle intervention with a repeat test? Do they need a liver biopsy? Do they need additional cross-sectional imaging to further raise your suspicion that this person has cirrhosis? So in general, I would apply these rules in the following way. The first is to try to enrich the population that we test with a higher pre-test probability of advanced fibrosis, which is not to select people willy-nilly, but to understand that it is those people with chronic metabolic disease, it is those people with a long history of alcohol use disorder, or nodular ultrasound that was detected incidentally, or outward signs of advanced liver disease on their skin or person. Once we have done that, we have done the work of sufficiently raising the pre-test probability so that these non-invasive tests can do the work of bringing us to a place of clinical certainty. The next step is to apply fibrosis 4 index. If you have a positive or an indeterminate result, then I would select based on whether or not I think the patient is going to have a successful fibroscan, and that means if they have a BMI less than 40, or if they have a BMI greater than 40, I would go for magnetic resonance elastography, and I would have a relatively higher threshold, a liver stiffness of 15 by fibroscan, or 5 by magnetic resonance elastography. For the average person who is 55 or greater, this might be enough for you to decide that they have cirrhosis, but if they are older, or they are sufficiently concerned and need additional tests to be confident to move forward, then additional tests like a liver biopsy might be needed. But the third step in this paradigm is to think about how we would change management for the patient on the basis of these non-invasive tests. And so we have to think about people within the continuum of risk, to look back at that paper by Shearer and Rowe, and think about the probability that they will decompensate or develop liver cancer in the next 5 to 10 years, and decide when we will intervene and with what. So someone who is relatively low risk, who just meets the bar for advanced fibrosis or cirrhosis according to conventional cut-offs, this is not somebody who is going to develop a decompensation. Someone with a liver stiffness of 10 to 12.5, and a fibrosis 4 of 2.66. But this is somebody who needs an intervention and now, relating to their lifestyle modification, abstinence from alcohol, or definitely treatment of their otherwise underlying liver diseases. As the liver stiffness starts to go up, as the platelet count starts to fall, as the fib 4 starts to go far beyond 2.67, we become increasingly worried about the risk of decompensation. And knowing what we know about our ability to prevent those events, we are starting to think immediately about pairing those point of care results with a discussion about whether or not we should be initiating non-selective beta blockers to blunt portal hypertension, or switch them from their selective beta blockers, or to something like metoprolol to carvedilol. So in summary, the most important thing to think about is just how confident we need to be in the performance of these tests. And I would argue for a 16-year-old, I don't really need to be confident, because I need to use these tools to motivate significant changes, not only in the patient, but the people that care for them, their health care providers, as well as their family. The tools, in general, are always going to be measures of things like the platelet count, fib 4, and elastography, tailored to whether or not you're going to get a viable result. The target is not staging, but rather the probability that that patient will get sick. And the application is to understand how we can assess the person with average risk, and then make a decision based on those results in the end. And I thank you for your attention.

non-invasive tests

liver disease management

probability of disease progression

personalized approach

FIB4 and elastography